Meta, Amazon, Walmart and Other Major Enterprises Publicly Restrict AI Service Usage Amid Cost Pressures

Companies that have actively promoted AI tools to their employees are now limiting usage as cost pressures continue to mount.

According to a recent Financial Times report, Amazon, Walmart, Cisco, Uber, and Meta are implementing usage restrictions to curtail wasteful consumption and encouraging the adoption of more cost-effective models to control AI spending.

The report indicates that these measures mark a new phase in enterprise AI adoption. As employees begin using AI agents capable of autonomously executing complex tasks beyond chatbots, the required computing resources have increased significantly. The report notes that enterprises are now in a position where they must weigh whether the tasks assigned to AI agents justify the investment.

As Anthropic and OpenAI shift portions of their services from flat-rate plans to token-based pricing models that charge per unit of data processed, the cost burden appears to intensify further. Enterprises may equip each employee with 10, 100, or even as many as 1,000 agents in more aggressive scenarios; if these agents run continuously, computing resource consumption will inevitably spike dramatically.

Software company Workato saw usage surge after approximately 1,300 employees began using AI agents starting last summer. In May, following Anthropic's transition to token-based pricing, the company's AI spending increased significantly.

Kosti Perikoc, Deloitte's Global Generative AI Leader, stated: "Computing costs are now beginning to enter the view of chief financial officers and boards of directors." He added: "Consumers and enterprises have been conditioned to believe AI is cheap or free, but the reality is far from it."

OpenAI CEO Sam Altman stated: "Cost has become a 'huge problem' for customers this year," adding: "This problem fundamentally did not exist last year."

Goldman Sachs analysts predicted last month that token consumption will grow 24-fold by 2030 due to AI agent usage, and that chip shortages will further deteriorate over the next 12 to 18 months as computing demand surges.